Changes Associated with Carbon Cycle

Climate change may affect the global biogeochemical cycles changing the magnitude of the natural sources and sinks of major GHGs. Numer-ous studies investigated the interactions between climate change and the carbon cycle (e.g., Friedlingstein et al., 2006), methane cycle (e.g., O’Connor et al., 2010), ozone (Cionni et al., 2011) or aerosols (e.g., Carslaw et al., 2010). Many CMIP5 ESMs now include a representa-tion of the carbon cycle as well as atmospheric chemistry, allowing interactive projections of GHGs (mainly CO2 and O3) and aerosols. With such models, projections account for the imposed changes in anthro-pogenic emissions, but also for changes in natural sources and sinks as they respond to changes in climate and atmospheric composition. If included in ESMs, the impact on projected concentration, RF and hence on climate can be quantified. Climate-induced changes on the carbon cycle are assessed below, while changes in natural emissions of CH4

are assessed in Chapter 6, changes in atmospheric chemistry in Chap-ter 11, and climate–aerosol inChap-teractions are assessed in ChapChap-ter 7.

12.4.8.1 Carbon Dioxide

As presented in Section 12.3, the CMIP5 experimental design includes, for the RCP8.5 scenario, experiments driven either by prescribed anthropogenic CO2 emissions or concentration. The historical and 21st century emission-driven simulations allow evaluating the cli-mate response of the Earth system when atmospheric CO2 and the cli-mate response are interactively being calculated by the ESMs. In such ESMs, the atmospheric CO2 is calculated as the difference between the imposed anthropogenic emissions and the sum of land and ocean carbon uptakes. As most of these ESMs account for land use changes and their CO2 emissions, the only external forcing is fossil fuel CO2

emissions (along with all non-CO2 forcings as in the C-driven RCP8.5 simulations). For a given ESM, the emission driven and concentration driven simulations would show different climate projections if the simulated atmospheric CO2 in the emission driven run is significantly different from the one prescribed for the concentration driven runs.

This would happen if the ESMs carbon cycle is different from the one simulated by MAGICC6, the model used to calculate the CMIP5 GHGs concentrations from the emissions for the four RCPs (Meinshausen et al., 2011c). When driven by CO2 concentration, the ESMs can calculate the fossil fuel CO2 emissions that would be compatible with the pre-scribed atmospheric CO2 trajectory, allowing comparison with the set of CO2 emissions initially estimated by the IAMs (Arora et al., 2011;

Jones et al., 2013) (see Section 6.4.3, Box 6.4).

Figure 12.36 shows the simulated atmospheric CO2 and global aver-age surface air temperature warming (relative to the 1986–2005 ref-erence period) for the RCP8.5 emission driven simulations from the CMIP5 ESMs, compared to the concentration driven simulations from the same models. Most (seven out of eleven) of the models estimate a larger CO2 concentration than the prescribed one. By 2100, the multi-model average CO2 concentration is 985 ± 97 ppm (full range 794 to 1142 ppm), while the CO2 concentration prescribed for the RCP8.5 is 936 ppm. Figure 12.36 also shows the range of atmospheric CO2

projections when the MAGICC6 model, used to provide the RCP con-centrations, is tuned to emulate combinations of climate sensitivity

uncertainty taken from 19 CMIP3 models and carbon cycle feedbacks uncertainty taken from 10 C⁴MIP models, generating 190 model simu-lations (Meinshausen et al., 2011c; Meinshausen et al., 2011b). The emulation of the CMIP3/C⁴MIP models shows for the RCP8.5, a range of simulated CO2 concentrations of 794 to 1149 ppm (90% confidence level), extremely similar to what is obtained with the CMIP5 ESMs, with atmospheric concentration as high as 1150 ppm by 2100, that is, more than 200 ppm above the prescribed CO2 concentration.

Global warming simulated by the E-driven runs show higher upper ends than when atmospheric CO2 concentration is prescribed. For the models assessed here, the global surface temperature change (2081–

2100 average relative to 1986–2005 average) ranges between 2.6°C and 4.7°C, with a multi-model average of 3.7°C ± 0.7°C for the con-centration driven simulations, while the emission driven simulations give a range of 2.5°C to 5.6°C, with a multi-model average of 3.9°C

± 0.9°C, that is, 5% larger than for the concentration driven runs. The models that simulate the largest CO2 concentration by 2100 have the largest warming amplification in the emission driven simulations, with an additional warming of more than 0.5°C.

The uncertainty on the carbon cycle has been shown to be of com-parable magnitude to the uncertainty arising from physical climate processes (Gregory et al., 2009). Huntingford et al. (2009) used a simple model to characterize the relative role of carbon cycle and climate sen-sitivity uncertainties in contributing to the range of future temperature changes, concluding that the range of carbon cycle processes represent about 40% of the physical feedbacks. Perturbed parameter ensembles systematically explore land carbon cycle parameter uncertainty and illustrate that a wide range of carbon cycle responses are consistent with the same underlying model structures and plausible parameter ranges (Booth et al., 2012; Lambert et al., 2012). Figure 12.37 shows how the comparable range of future climate change (SRES A1B) arises from parametric uncertainty in land carbon cycle and atmospheric feedbacks. The same ensemble shows that the range of atmospheric CO2 in the land carbon cycle ensemble is wider than the full SRES con-centration range (B1 to A1FI scenario).

The CMIP5 ESMs described above do not include the positive feed-back arising from the carbon release from high latitudes permafrost thawing under a warming scenario, which could further increase the atmospheric CO2 concentration and the warming. Two recent studies investigated the climate–permafrost feedback from simulations with models of intermediate complexity (EMICs) that accounts for a per-mafrost carbon module (MacDougall et al., 2012; Schneider von Deim-ling et al., 2012). Burke et al. (2012) also estimated carbon loss from permafrost, from a diagnostic of the present-day permafrost carbon store and future soil warming as simulated by CMIP5 models. However, this last study did not quantify the effect on global temperature. Each of these studies found that the range of additional warming due to the permafrost carbon loss is quite large, because of uncertainties in future high latitude soil warming, amount of carbon stored in permafrost soils, vulnerability of freshly thawed organic material, the proportion of soil carbon that might be emitted as carbon dioxide via aerobic decomposition or as methane via anaerobic decomposition (Schneider von Deimling et al., 2012). For the RCP8.5, the additional warming from permafrost ranges between 0.04°C and 0.69°C by 2100 although

12

there is medium confidence in these numbers as are the ones on the amount of carbon released (see Section 12.5.5.4) (MacDougall et al., 2012; Schneider von Deimling et al., 2012).

12.4.8.2 Changes in Vegetation Cover

Vegetation cover can also be affected by climate change, with forest cover potentially being decreasing (e.g., in the tropics) or increasing (e.g., in high latitudes). In particular, the Amazon forest has been the subject of several studies, generally agreeing that future climate change would increase the risk tropical Amazon forest being replaced by seasonal forest or even savannah (Huntingford et al., 2008; Jones et al., 2009; Malhi et al., 2009). Increase in atmospheric CO2 would partly reduce such risk, through increase in water efficiency under ele-vated CO2 (Lapola et al., 2009; Malhi et al., 2009). Recent multi-model estimates based on different CMIP3 climate scenarios and different dynamic global vegetation models predict a moderate risk of tropical forest reduction in South America and even lower risk for African and Asian tropical forests (see also Section 12.5.5.6) (Gumpenberger et al., 2010; Huntingford et al., 2013).

200 300 400 500 600 700 800 900 1000 1100 1200

CO2 concentration (ppm)

a Atmospheric CO₂ concentration

c Atmospheric CO₂ concentration

−1 0 1 2 3 4 5 6 7

b Global mean surface air temperature

d Global mean surface air temperature

1850 1900 1950 2000 2050 2100

200 300 400 500 600 700 800 900 1000 1100

1850 1900 1950 2000 2050 2100

−1 0 1 2 3 4 5 6

Global−Mean Temperature relative to 1986−2005 (°C)

CMIP3 & C4MIP emulation:

CMIP5: CMIP5:

CMIP3 & C4MIP emulation:

90%

68%50%Ranges

Concentration-driven default

90%

68%50%Ranges Concentration-driven default

Emission-driven Emission-driven

Concentration-driven

C C

Figure 12.36 | Simulated changes in (a) atmospheric CO2 concentration and (b) global averaged surface temperature (°C) as calculated by the CMIP5 Earth System Models (ESMs) for the RCP8.5 scenario when CO2 emissions are prescribed to the ESMs as external forcing (blue). Also shown (b, in red) is the simulated warming from the same ESMs when directly forced by atmospheric CO2 concentration (a, red white line). Panels (c) and (d) show the range of CO2 concentrations and global average surface temperature change simulated by the Model for the Assessment of Greenhouse Gas-Induced Climate Change 6 (MAGICC6) simple climate model when emulating the CMIP3 models climate sensitivity range and the Coupled Climate Carbon Cycle Model Intercomparison Project (C⁴MIP) models carbon cycle feedbacks. The default line in (c) is identical to the one in (a).

Figure 12.37 | Uncertainty in global mean temperature from Met Office Hadley Centre climate prediction model 3 (HadCM3) results exploring atmospheric physics and ter-restrial carbon cycle parameter perturbations under the SRES A1B scenario (Murphy et al., 2004; Booth et al., 2012). Relative uncertainties in the Perturbed Carbon Cycle (PCC, green plume) and Perturbed Atmospheric Processes (PAP, blue plume) on global mean anomalies of temperature (relative to the 1986–2005 period). The standard simulations from the two ensembles, HadCM3 (blue solid) and HadCM3C (green solid) are also shown. Three bars are shown on the right illustrating the 2100 temperature anomalies associated with the CMIP3/AR4 ensemble (black) the PAP ensemble (blue) and PCC ensemble (green). The ranges indicate the full range, 10th to 90th, 25th to 75th and 50th percentiles.

CMIP 3PAP PCC

12

Figure 12.38 | Impact of land use change on surface temperature. LUCID-CMIP5 experiments where six ESMs were forced either with or without land use change beyond 2005 under the RCP8.5 scenario. Left maps of changes in total crop and pasture fraction (%) in the RCP8.5 simulations between 2006 and 2100 as implemented in each ESM. Right maps show the differences in surface air temperature (averaged over the 2071–2100 period) between the simulations with and without land use change beyond 2005. Only statistically significant changes (p < 0.05) are shown.

Difference in crop and pasture fraction (%) Change in surface air temperature (°C)

12

ESMs simulations with interactive vegetation confirmed known bio-physical feedback associated with large-scale changes in vegetation.

In the northern high latitudes, warming-induced vegetation expansion reduces surface albedo, enhancing the warming over these regions (Falloon et al., 2012; Port et al., 2012), with potentially larger ampli-fication due to ocean and sea ice response (Swann et al., 2010). Over tropical forest, reduction of forest coverage would reduce evapotran-spiration, also leading to a regional warming (Falloon et al., 2012; Port et al., 2012).

CMIP5 ESMs also include human induced land cover changes (deforest-ation, reforestation) affecting the climate system through changes in land surface physical properties (Hurtt et al., 2011). Future changes in land cover will have an impact on the climate system through bio-physical and biogeochemical processes (e.g., Pongratz et al., 2010).

Biophysical processes include changes in surface albedo and changes in partitioning between latent and sensible heat, while biogeochemi-cal feedbacks essentially include change in CO2 sources and sinks but could potentially also include changes in N2O or CH4 emissions. The bio-physical response to future land cover changes has been investigated within the SRES scenarios. Using the SRES A2 2100 land cover, Davin et al. (2007) simulated a global cooling of 0.14 K relatively to a simulation with present-day land cover, the cooling being largely driven by change in albedo. Regional analyses have been performed in order to quantify the biophysical impact of biofuels plantation generally finding a local to regional cooling when annual crops are replaced by bioenergy crops, such as sugar cane (Georgescu et al., 2011; Loarie et al., 2011). How-ever, some energy crops require nitrogen inputs for their production, leading inevitably to nitrous oxide (N2O) emissions, potentially reduc-ing the direct coolreduc-ing effect and the benefit of biofuels as an alterna-tive to fossil fuel emissions. Such emission estimates are still uncertain, varying strongly for different crops, management methods, soil types and reference systems (St. Clair et al., 2008; Smeets et al., 2009).

In the context of the Land-Use and Climate, IDentification of robust impacts (LUCID) project (Pitman et al., 2009) ESMs performed addi-tional CMIP5 simulations in order to separate the biophysical from the biogeochemical effects of land use changes in the RCP scenarios.

The LUCID–CMIP5 experiments were designed to complement RCP8.5 and RCP2.6 simulations of CMIP5, both of which showing an intensi-fication of land use change over the 21st century. The LUCID–CMIP5 analysis was focussed on a difference in climate and land-atmosphere fluxes between the average of ensemble of simulations with and with-out land use changes by the end of 21st century (Brovkin et al., 2013).

Due to different interpretation of land use classes, areas of crops and pastures were specific for each ESM (Figure 12.38, left). On the global scale, simulated biophysical effects of land use changes projected in the CMIP5 experiments with prescribed CO2 concentrations were not significant. However, these effects were significant for regions with land use changes >10%. Only three out of six participating models, CanESM2, HadGEM2-ES and MIROC-ESM, reveal statistically signifi-cant changes in regional mean annual mean surface air temperature for the RCP8.5 scenario (Figure 12.38, right). However, there is low confidence on the overall effect as there is no agreement among the models on the sign of the global average temperature change due to the biophysical effects of land use changes (Brovkin et al., 2013).

Changes in land surface albedo, available energy, latent and sensible

heat fluxes were relatively small but significant in most of ESMs for regions with substantial land use changes. The scale of climatic effects reflects a small magnitude of land use changes in both the RCP2.6 and 8.5 scenarios and their limitation mainly to the tropical and subtropical regions where differences between biophysical effects of forests and grasslands are less pronounced than in mid- and high latitudes. LUCID-CMIP5 did not perform similar simulations for the RCP4.5 or RCP6.0 scenarios. As these two scenarios show a global decrease of land use area, one might expect their climatic impact to be different from the one seen in the RC2.6 and RCP8.5.

12.4.9 Consistency and Main Differences Between